Mechanistic interpretability provides white-box analysis methods to inspect whether artificial neural networks possess internal causal structures equivalent to conscious minds